Defects in SiO2 in buried-oxide structures formed by O+ implantation

433

Nuclear Instruments and Methods in Physics Research B32 (1988) 433-436 North-Holland, Amsterdam

DEFECTS

IN SiO, IN BURIED-OXIDE

R.C. BARKLIE

‘) , T.J. ENNIS

STRUCTURES

*), P.L.F. HEMMENT

FORMED BY O+ IMPLANTATION

2, and K. REESON

” Physrcs Department, Trinrty College, Dublin 2, Ireland 2, Department of Electronic and EIectricaI Engineermg. University of Surrey, Gurldjord

2).

UK

EPR measurements have been made on (100) n type silicon wafers implanted with 200 keV t60+ ions in the dose range (0.5-1.4)X lOi* 0+ cme2 and at implantation temperatures, T,, in the range 250-600 o C. Several types of defect are observed including E; centres. For a dose of 1.4~ lo’* 0 cm-* the E; concentration is found to increase from 1.5 x 1014 to 2.3 X lOI cmm2 as T, increases from 250 to 6OO“C. Measurements on lower dose samples indicate that E; defects are present in SiO, precipitates as well as in the buried SiO, layer. The E; EPR signal only starts to decrease for annealing temperatures S 450 o C.

2. Experimental details

1. Introduction Ion beam synthesis, being

used

(SOI)

structures

parameters controls

to make

most often with oxygen high

quality

[1,2]. One of the most

is the implantation

the level of radiation

ions, is

silicon-on-insulator

temperature

important T,. This

damage and hence whether

or not any amorphous silicon layers form [3,4]; it also determines the extent of oxide nucleation and precipitation as these are temperature dependent processes [5,6]. Previous experiments [7,8] have identified some of the defects present in these SO1 structures. The aims of the present work are to investigate the effect of changing T, on the concentration, and possibly the type, of defects present and to find in each case the extent to which the defects can be removed by annealing. In this paper we concentrate on the defects present within the SiO, precipitates including the buried layer.

(100) n type silicon wafers were implanted at the University of Surrey, under the conditions listed in table 1. The implantation temperature Ti was varied by varying the beam current since conductive heat losses were minimized by supporting the sample on silicon tips. As only the central 2.5 X 2.5 cm section of each 3 in. wafer was implanted a temperature difference of about 20 o C existed between the centre and edge of the implanted region [9]. A dose of 1.4 X lo’* 0 cme2 has been found, for 200 keV O+, to be the critical value, &,, required for the peak oxygen concentration to reach that corresponding to stoichiometric SiO, [lo]. Detailed results for sample 167 have been given elsewhere [7]; it has a buried oxide of thickness = 0.28 pm which is stoichiometric and amorphous, and both the top silicon layer of thickness

Table 1 Sample implantation conditions. The astensk mdicates equivalent O+ values Sample

Ion

Beam current (PA)

Implantation temperature

Dose (lo’* cmm2)

(keV)

Ion energy

1.4 1.4 1.4 1.4 1.8* 0.5 0.75 0.95

200 200 200 200 200* 200 200 200

(“C) S020a SO2la S023a S033a 167 SO29 SO36 so31

160+ 16o+ 160+ 160+ 32 0; 160+ 160+ MO+

0168-583X/88/$03.50 (North-Holland

Physics

50*1 70*1 90*1 123+3 50&2 llOf2 117f4 119+1

250 350 450 600 520 540 570 580

0 Elsevier Science Publishers B.V. Publishing

Division)

VIII. DEVELOPING

TRENDS

434

R.C. Barklie et al. / Defects in SO, in buned oxrde structures

= 0.4 pm and the layers below it are crystalline. However, for T, = 4QO* C amorphous silicon layers are present in the as-implanted state 1113. EPR measurements were made at Trinity College using a Bruker spectrometer with 100 kHz field modulation. The measurements were made at room temperature and with a microwave frequency of about 9.6 GHz.

Figs. 1 a-d show the effect on the EPR spectrum, for magnetic field B tt [lOOI, of changing the implantation temperature Ti in the range 250-600°C for a constant dose, 1.4 X 1018 O+ cme2, of 200 keV I%‘+, ias; the sampfcs are ia the as-imphmted state. Also included, in fig. le, is the previously reported j?] spectrum for sample 167. Two features A and G characterise each spectrum. A was shown [7] to be a superposition of two spectra arising from Pa like centres, predominantly at the Si/SiO, interfaces of SiOZ precipitates, and amorphous silicon centres which may be associated with oxygen. In this paper we concentrate on feature C. it saturates relatively easily with microwave power and for each of the spectra in figs. 1 a-d it can be fitted to a slightly asymmetric, but isotropic, line with a zero crossing g value of 2.0006 f 0.0004 and peak-to-peak width A fspp = 0.33 Jr 0.02 mT. The corresponding values found f7] for sample 167 were g = 2.0003 f 0.0003 and AI&, = 0.20 f 0.02 mT. These g values are similar to 2.0002

’

1mT

B-

Fis_ 1. BPR specna, for itfl flt?0] of samples (a) SQ33a fb) S023a (c) S021a (d) S02Qa and (e) 167 in their as-implanted states. Table 1 shows their implantation conditions

4 0.0005 reported [I23 for El centres in ion implanted amorphous SiO, and, as we suggested before [7] for sample 167, it is very likely that feature C is due to E; centres in amorphous SiO,. The broadened lineshape is symptomatic of a distribution of g values but may also arise from the spin-spin interactions which can contribute to the broadening [12]. Figs. 1 a-d show that the ratio (intensity of feature A)/~inte~ity of feature C) decreases as T increases. From a comparison of the intensities of computed fits to these features with that for a standard sample of pitch in KC1 we obtain the defect concentrations N(A) = 1.5 X 1015, 1.4 X 1015, 1.2 X 1015, 1.1 X 1015 cm-‘, N(C) = 1.5 x 1014, 2.0 x 10i4, 2.0 x 10x4, 2.3 x 1Or4 cm-z for T, = 250, 350, 454 600 o C, respectively. In evaluating N account was taken of changes in the cavity Q value and filling factor. The decrease in N(A) as T, increases might be expected since it is known that amorphous silicon layers are present at the lower temperatures but not at 6oO°C [ll]. What is surprising is the increase in the E; centre concentration as T, increases. It is surprising since for ion implanted samples of amorphous Sicl, it has been found that the El EPR signal starts to fall at an anneal temperature of about 100°C [12]. The presence of E,’ centres even though T, 39 100°C probably occurs because the ion beam simultaneously creates Ei centres as well as annealing them, but a higher concentration (per unit area) at higher T, v&szs is unexpected. It seems reasonable to suppose however, that an oxide precipitate must exceed a certain size and possibly have a suitable shape to be able to contain E,/ centres. Can for example, the ribbon like precipitates seen after heating Czochralski grown silicon at 650 DC [6] contain Ei centres? As growth is enhanced by raising the temperature [5]* it may there fore be that as T, is increased so too is the volume fraction of oxide precipitates which can contain Ei centres. The above explanation assumes that E; centres can occur in SiO, precipitates. To check this we recorded EPR spectra for sampIes 5029, SO36 and SO31 which, as shown in table 1, have doses less than & and hence have no buried layer in the as-implanted state. Fig. 2 shows that feature C is present for two of these samples and computed fits yield zero crossing g values of 2.0004, 2.0005 for figures 2b, c respectively which justifies the above assumption. The fits yield defect ~n~n~a~o~s N(E;) = 0.19 X 10r4, ‘1.4 X 10f4 cm-’ for doses of 0.75 X I@ 0.95 X 10” O+ cm-* respectively. It is’ interesting to compare the E,’ concentration per unit volume of SiO, precipitate (including buried layer) for samples S033a and SO31 both of which were implanted at about the same temperature. For a sample implanted with 1.8 X 10’s U+ CM-~, 200 keV Ut, the oxygen concentration even at the top surface of the upper silicon layer was as high as 2 X 1CP atoms cmd3

R.C. Barklie et al. / Defects in SQ

n v

’

DOSE ( 1 d”dcm*,

1mT

I

Fig. 2. The dose dependence of the EPR signal, for B II [lOO], for samples (a) SO29 (b) SO36 and (c) SO31 in the as-implanted state.

[lo]. This so much exceeds the solubility of oxygen in silicon of - 1018 atoms cmp3 that it is reasonable to assume that all the oxygen for samples S033a and SO31 is precipitated. Since for stoichiometric SiO, there are 4.4 x 1O22 oxygen atoms cmp3, this implies that the volume of oxide precipitates (including any buried layer) per unit area of sample surface is 2.2 X 10e5 and 3.2 X 10m5 cm3 for doses of 0.95 X 10” and 1.4 X 101* O+ cmp2, respectively, and hence the corresponding E,’ concentrations are 6.4 X 10” and 7.2 X 10” cmm3. The similarity of these values suggests that the E,’ concentration has reached a constant value and that changes in E,’ population with dose just reflect changes in the total volume of SiO, precipitates. However, on this basis, larger E; EPR signals would be expected both for sample S029, for which no signal at all is seen, and also for sample SO36, so the above similarity is probably just coincidental. The concentrations of 6.4 x lOI and 7.2 x 10” crnp3 are surprisingly close to the value of the order of 1.5 x 1019 cmm3 at which the concentration saturates in amorphous SiO, ion implanted at room temperature [13] and only an order of magnitude less than the value of 6 x 10” cme3 estimated for an SiO, film irradiated with electrons to a dose of 4 x 1Ol2 rad (Si) [14]. Indeed, if, as suggested earlier, not all oxide precipitates contain E,’ centres the local concentrations could be even higher. One final point to note about the data in fig. 1 is that the ratio (intensity of feature A)/(intensity of feature C) for sample 167 is greater than for the other four samples. This is mainly due to the smaller value of 0.7 x 1014 cmp2 found for the E; concentration in this sample [7]. The only obvious differences between the

435

m buried oxide structures

implant conditions for this sample and those of the others are the use of 3202+ rather than 160+ and the higher dose of 1.8 x 1Ol8 0 cme2. However, Golanski et al [15] found no difference in Ei production when using atomic or molecular oxygen beams and the higher dose, leading as it does to a greater volume of SiO,, would be expected to produce a higher Ei concentration (per unit area) so the observed lower concentration is puzzling. Each of the samples S020a, S023a, S033a and 167 has been annealed for 10 min intervals at successively higher temperatures from room temperature up to about 1100 o C in 50 o C steps. Fig. 3 shows that the intensity of the E; spectrum starts to fall at about 450” C and 350 OC for the 160+ samples and sample 167 respectively. This annealing behaviour differs from that observed for E,’ centres created by ion implantation into thermally grown SiO, where, although irreversible annealing only sets in above 500” C [15,16], reversible annealing begins at about 100 o C [12]. Various processes have been suggested for the low temperature ( ( 400 ’ C) annealing including hole release to form the diamagnetic B, centre [17], combination with small neutral molecules such as H,O [18] or combination with 0, to form peroxy radicals 1191. We were unable to detect the formation of peroxy radicals and molecules such as H,O should be absent from the buried oxide so the

1-o

O-8

0

0.6

0

0 -r,

0” 8 V

o-4 0

o” o-2

B q

0

200

400

800

T(k) Fig. 3. The intensity, I, of the E; EPR signal as a function of the anneal temperature, T. I = I, for the as-implanted state of samples S033a Co), S023a (v), S020a (0) and 167 (0) implanted at 600, 450, 250 and 520 o C respectively. VIII. DEVELOPING TRENDS

436

R. C. Barkhe ei al. / Defects in SO,

mechanism operating in our samples is unclear. We hope to clarify the situation by finding out whether the observed annealing is reversible or not.

4. Conclusions Samples implanted with 1.4 X lo’* 0+ cmp2 using 200 keV “O+ ions have an E,’ defect concentration in the as-implanted state which increases from 1.5 x 1014 to 2.3 x 1014 cmw2 as q is raised from 250 to 600 o C. This increase is attributed to an increase in the volume of SiO, available for E,’ production. From studying lower dose samples it is shown that E,’ centres can exist in SiO, precipitates as distinct from buried layers. The E,’ EPR signal, in samples implanted with 160+, only starts to decrease at an anneal temperature of about 450 o C and becomes unobservable at about 600 o C.

References

PI K. Izumi, Y. Omura and S. Nakashima, Mat. Res. SK Symp. Proc. (EIsevier, Amsterdam, 1984) vol. 23, p. 443. Materials Research Society, Fall Meeting, Boston, (Dec. 1985). [31 R.F. Pinirzotto, J. Cryst. Growth 63 (1983) 559.

PI P.L.F. Hemment,

in buried oxide structures

[4] C.G. Tuppen, M.R. Taylor, P.L.F. Hemment and R.P. Arrowsmith, Appl. Phys. Lett. 45 (1984) 57. [5] R.C. Newman, M.J. Binns, W.P. Brown, F.M. Livingston, S. Messoloras, R.J. Stewart and J.G. Wilkes, Physica 116B (1983) 264. [6] A. Bourret, J. Thibault-Desseaux and D.N. Seidman, J. Appl. Phys. 55 (1984) 825. [7] R.C. Barkhe, A. Hobbs, P.L.F. Hemment and K. Reeson, J. Phys. Cl9 (1986) 6417. [8] T. Makino and J. Takahashi, Appl. Phys. Lett. 50 (1987) 267. [9] J.R. Davis, M.R. Taylor, G.D.T. Spiller, P.J. Skevington and P.L.F. Hemment, Appl. Phys. Lett. 48 (1986) 1279. [lo] P.L.F. Hemment, E. Maydell-Ondrusz, K.G. Stephens, J. Butcher, D. Ioannou and J. Alderman, Nucl. Instr. and Meth. 209/210 (1983) 157. [ll] C.G. Tuppen, M.R. Taylor, P.L.F. Hemment and R.P. Arrowsmith, Appl. Phys. Lett. 45 (1984) 57. [12] R.A.B. Devine, Nucl. Instr. and Meth. Bl (1984) 378. [13] R.A.B. Devine and A. Golanski, J. Appl Phys. 54 (1983) 3833. [14] R.L. Pfeffer, J. Appl. Phys. 57 (1985) 5176. [15] A. Golanski, R.A.B. Devine and J.C. Oberlin, J. Appl. Phys. 56 (1984) 1572. [16] R.A.B. Devine, J. Appl. Phys. 56 (1984) 563. [17] M. Antonini, P. Camagni, P.N. Gibson and A. Manara, Radiat. Eff. 65 (1982) 49. [18] D.L. Griscom, J. Non-Cryst. Solids 73 (1985) 51. [19] A.H. Edwards and W.B. Fowler, Phys. Rev. B26 (1982) 6649.